Life In 19x19
http://lifein19x19.com/

Laerthd study journal
http://lifein19x19.com/viewtopic.php?f=48&t=13412		 Page 7 of 8
Author:  Laerthd [ Wed May 23, 2018 3:57 am ]
Post subject:  Re: Laerthd study journal
Hi Ed,

Thanks for the variations!

:b29:, I did not realize how tight white would be with a move like this.
I thought of playing maybe B11 but I couldn't see a clear path.
Just putting pressure without trying to kill seems much better.

:w46: this seems quite similar to me. what is the advantage of playing O2 directly?
If anything, as white I would play P4 and next play toward the center with :w48: at L5

:b55: yeah I was not sure about this sequence so I played it safe

For the AI analysis, I tried using Lizzie but it would start so I will have to look more into this.
Author:  EdLee [ Wed May 23, 2018 11:16 am ]
Post subject: 
Hi Laerthd,

:w46: at o2: W still has P4 atari ( too slow for B to play there now ),
and aji of R3 peep, R2 clamp, etc.
Author:  Laerthd [ Wed May 23, 2018 1:40 pm ]
Post subject:  Re: Laerthd study journal
Hi Ed,

That make sense, thanks!


Still haven't succeded to use Lizzie but I used go review partner with AQ,

To answer my question for :w6:, apparently pincering is a little inferior to just play F17 (49.14% Vs 50.02%)

Click Here To Show Diagram Code
[go]$$B
$$ +---------------------------------------+
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . 2 . . . . . . . . . O 7 . . |
$$ | . . . O . . . . . , . . . 6 . , . . . |
$$ | . . . . . . . . . . . . . . . . 3 . . |
$$ | . . X . . . . . . . . . . . 5 . . . . |
$$ | . . . . . . . . . . . . . . . 4 . . . |
$$ | . . O . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . , . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . 1 . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . X . . . |
$$ | . . . X . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ +---------------------------------------+[/go]


As Ed noted, failing to keep pressure on white group caused a big drop in my winning probability
:b29: at c6 is valued at 50.9% and 47% for R14

I am not sure how to interpret these numbers however since for :w46: AQ would play O2 but gives a better chance of winning for P4

My big mistake was :b93: dropping my winning probability from 55.52% to 32.42%.
Here is what AQ would do :
Click Here To Show Diagram Code
[go]$$B
$$ +---------------------------------------+
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . X X O . . . . . . . . . . O X . . |
$$ | . . X O . . . . . . . O . . . O X . . |
$$ | . . X O . O O . . , . . . . O O X . . |
$$ | . . X O . X X O O . . . . . O X X . . |
$$ | . . X O . . . X . . . . . . . O X . . |
$$ | . . O X X . . X . O . . . O . O X . . |
$$ | . . O . . . . . . . . . . O X X . . . |
$$ | . . . O X . . . . . . . . O . X X . . |
$$ | . . O X X . . . . , . . . . O X O . . |
$$ | . . O . . . . . . . . . . . O X O . . |
$$ | . . . . . . . . . . . . . . . O O . . |
$$ | . . . O . . . . . . . . X O O X O . . |
$$ | . . . O X . . . . . 3 . . X X X X . . |
$$ | . . O X X . . . . X 2 4 . . . . . . . |
$$ | . O O , . . . . . 7 8 . . O O X . . . |
$$ | . . X X . X . . X . O . . O X X . . . |
$$ | . . . . . . . . . 1 5 6 . O O X . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ +---------------------------------------+[/go]
Author:  EdLee [ Wed May 23, 2018 2:10 pm ]
Post subject: 
Hi Laerthd,

Thanks. Is AQ running on your home computer, Windows or Mac ?
Author:  Laerthd [ Wed May 23, 2018 2:22 pm ]
Post subject:  Re: Laerthd study journal
On a laptop running windows.
Probably not the best analysis tool but surely good enough for my level!
Author:  Laerthd [ Fri May 25, 2018 3:48 pm ]
Post subject:  Re: Laerthd study journal
Here are 3 games I played this week, 8k on Fox Go.
I ran go review partner with AQ on these games and it looks useful but only after a self-review, to see my blind spots. As daal noted on another topic, go is still missing quite a lot in term of quality review tools.
I would attach the win rates at each move but I cannot find how to include a picture from my hard drive.

Main insight : I did not realize how bad it was to mess up the upper right, AQ says -57.66% of win progability (from 75.62% to 17.96%)


Main insights:
I did not see how big it was for my top right group to expand to the upper side (:b67: at P18) as well as how I could make it live in the corner and avoid the reduction of my lower right corner


Main insights:
I did not realize how weak the 2nd line stones on the top were
I did not evaluate correctly how much I could get from the right side, much less than expected



Attachments:
[laerth]vs[ihnqke]QuickReview.sgf [1.55 KiB]
Downloaded 805 times
[laerth]vs[门斯克]Review.sgf [3.72 KiB]
Downloaded 787 times
[laerth]vs[12.17]Review.sgf [4.36 KiB]
Downloaded 788 times
Author:  EdLee [ Fri May 25, 2018 7:31 pm ]
Post subject: 
Hi Laerthd,
Quote:
how to include a picture from my hard drive
Click "Browse...", navigate to your photo on your computer,
then click "Add the file":
Attachment:
x.jng.png
x.jng.png [ 21.16 KiB | Viewed 19519 times ]
After that, click "Place inline", like this:
Attachment:
y.png
y.png [ 8.46 KiB | Viewed 19519 times ]
Author:  Laerthd [ Sat May 26, 2018 3:34 am ]
Post subject:  Re: Laerthd study journal
Hi Ed,

I tried that but it wouldn't work. I just noticed the error message I get, 3 is the maximum number of attached files.

So here are the win rates:

Attachments:
[laerth]vs[ihnqke]WinRate.png
[laerth]vs[ihnqke]WinRate.png [ 10.43 KiB | Viewed 19506 times ]
[laerth]vs[门斯克]WinRate.png
[laerth]vs[门斯克]WinRate.png [ 18.22 KiB | Viewed 19506 times ]
[laerth]vs[12.17]WinRate.png
[laerth]vs[12.17]WinRate.png [ 14.58 KiB | Viewed 19506 times ]
Author:  EdLee [ Sat May 26, 2018 4:58 am ]
Post subject: 
Hi Laerthd,

But it worked -- you attached 3 photos. :)
Author:  Laerthd [ Sat May 26, 2018 11:35 am ]
Post subject:  Re: Laerthd study journal
Sorry I wasn't clear, I meant attach it to the first post, linked to the games.
Author:  EdLee [ Sat May 26, 2018 11:43 am ]
Post subject: 
Ah, ( 1 SGF + 1 photo ) per post. :)
Author:  Laerthd [ Sat Jun 02, 2018 2:33 am ]
Post subject:  Re: Laerthd study journal
I've noticed that I play better than usual since I started posting here again!

I reduced my games to 1 a day on average I would say. It really forces me to think so I don't lose the only game I'll have today.
The popular advice to improve is "just play games" but actually I would say "play fewer games" is a better advice.
Also posting here and analyzing my games with go review partner forces me to think about my moves.
Here because it doesn't do any good if for 3 games in a row the advice is "don't push if you can't cut".
For GRP, it is so frustrating to analyze a game to look for better alternative and from move 50 to 150 it tells you that literally every move you played was worse than this specific spot. On the other hand, it is great for finding the urgent points.

Here is a game for today that I am quite proud in which I play white. I did some errors in the lower side but I got what I thought was a not so bad result in the upper left. However, AQ disagrees and I don't know why. It gives me a <20% chance of winning after the corner sequence. Any comments on that part would be helpful.


Attachments:
Winning probability.PNG
Winning probability.PNG [ 13.8 KiB | Viewed 19384 times ]
[王柏泉]vs[laerth]Review.sgf [3.79 KiB]
Downloaded 797 times
Author:  EdLee [ Sat Jun 02, 2018 3:03 am ]
Post subject: 
Hi Laerthd,
Quote:
It gives me a <20% chance of winning after the corner sequence.
My guess is AQ thinks if the ladder is bad for :white:,
on your board, you shouldn't pick avalanche. You can easily test this. :)
See if :w10: is your last chance.

( Caveat: just because AQ/AZ thinks <20% winrate,
it doesn't necessarily mean (top) humans can take advantage of it. )
Author:  dfan [ Sat Jun 02, 2018 3:26 am ]
Post subject:  Re: Laerthd study journal
Laerthd wrote:
I reduced my games to 1 a day on average I would say. It really forces me to think so I don't lose the only game I'll have today.
The popular advice to improve is "just play games" but actually I would say "play fewer games" is a better advice.
The best advice is different for everyone! I need to play more games, but that's because I play fewer than one game a day. :)

Quote:
Here is a game for today that I am quite proud in which I play white. I did some errors in the lower side but I got what I thought was a not so bad result in the upper left. However, AQ disagrees and I don't know why. It gives me a <20% chance of winning after the corner sequence. Any comments on that part would be helpful.
With move 16 you give Black a very useful move (and a prisoner) completely for free. One plays this sort of move to cause shortage of liberties so you can squeeze or capture, but neither is true here. Moreover, you're in the middle of a big fight with two weak groups on either side, so effectively passing hurts even more. Maybe it only loses 10 points or so, but that's a huge difference for an engine that is computing the probability that Black will win a game between two very strong players starting from this position.
Author:  Knotwilg [ Sat Jun 02, 2018 5:01 am ]
Post subject:  Re: Laerthd study journal
On White 16
Author:  Laerthd [ Sat Jun 02, 2018 5:33 am ]
Post subject:  Re: Laerthd study journal
Ed,
no, up to :w15: AQ thinks I am doing ok

dfan,
Indeed when I played it I didn't know where it would go but I thought I could try some kind of squeeze.

Knotwilg,
Thanks for the explanation, it makes sense why this is bad for white
Author:  EdLee [ Sat Jun 02, 2018 6:36 pm ]
Post subject: 
Hi Laerthd, Thanks. Good to know.

I wonder if DM's AG teaching tool has your exact position up to the choice of avalanche.
Author:  Laerthd [ Wed Jun 06, 2018 4:23 am ]
Post subject:  Re: Laerthd study journal
Indeed Ed, it's here.

At move 15
Surprisingly, it gives 40.3% to F16 while AQ judge it at 50.78%.
AlphaGo prefers to extend F17 at 45.1% while AQ thinks it is worst than F16 (but still overvalue this position) at 49,16%

At move 16, once again their opinion differ.
AlphaGo move is B16 with 39.1% of winning probability for black
AQ chooses F15 first and B16 as the second choice but B16 is rated 53.08% for black!

That is a bit of a shame that such variations exist.
I had hoped that a system similar to chess analysis would be possible. You would get a rating of your move from slightly non-optimal to game-ending mistake but with this much difference in the evaluation, it is probably not for tomorrow.
Author:  Uberdude [ Wed Jun 06, 2018 4:28 am ]
Post subject:  Re: Laerthd study journal
Laerthd, I actually studied this small avalanche a bit with AG, the position of bottom right stone is very important for a ladder: forum/viewtopic.php?p=227181#p227181. I didn't check what AQ says, but check it understands the ladder; also what AG thinks is "best of a bad job" is not the usual human thinking in the avalanche joseki, maybe AQ is still more following the human ideas.
Author:  Laerthd [ Thu Jun 07, 2018 2:24 pm ]
Post subject:  Re: Laerthd study journal
I'll have a look, thanks uberdude.

In the meantime, I rose to 6k on foxgo and I am still about 50% win so I am quite happy about that.

Here are 2 games that I played today, one won, one lost.

The lost game was because of move 55 I think.
The won one I have no real question, I am just proud of how it went so I share it.





Attachments:
[火箭sjhs]vs[laerth]QuickLook.sgf [1.56 KiB]
Downloaded 733 times
[laerth]vs[nanhai123]Review.sgf [2.53 KiB]
Downloaded 746 times
Page 7 of 8 All times are UTC - 8 hours [ DST ]
Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group
http://www.phpbb.com/